Numerical Modeling of Liquefaction-induced Settlement due to Reconsolidation

Abstract

Soil liquefaction poses a major threat to human life and property. Settlement following liquefaction has been one of the major sources of liquefaction-induced damage in past earthquakes and creating resilient infrastructure requires methods to predict this settlement for various types of soil and site conditions. Empirical methods are commonly utilized in practice to evaluate settlements. However, these methods have some inherent limitations that might hinder the accuracy of settlement predictions for specific soils and site conditions. Numerical models are a may be used to predict post-liquefaction responses of soil while accounting for the complexities encountered in the field, such as variable stratigraphy, partial drainage, and soil-structure interaction. In order to have confidence in the results of these numerical models, they must first be validated using results from physical models or well-documented case histories. Another important factor in evaluating liquefaction-induced damage is the effect of spatial variability in soil properties. Many previous studies have focused on the response of soil layers with uniform properties, but this is a simplification of the true variability encountered in the field. It is not well-understood how spatial variability in soil properties affects reconsolidation settlements, so a more in-depth assessment is required. The primary motivation for this study was to improve the existing numerical protocols for liquefaction modeling in order to accurately predict observed liquefaction responses for a range of soil types and explore important factors that influence the magnitude and distribution of reconsolidation settlements. Previous studies in this area have primarily focused on a single site or a single type of soil. This precludes the ability to examine how soil type influences settlement patterns or to assess whether a numerical protocol can predict accurate settlements under various loading paths. This study fills this gap by applying a single numerical protocol (the numerical platform Fast Lagrangian Analysis of Continua, FLAC, and constitutive model PM4Sand) to model both excess pore pressure generation and dissipation for three types of problems (uniform centrifuge and shake table tests, centrifuge tests with a retaining wall, and a spatially variable field site) with six different types of soil. The necessity of soil-specific calibration of post-liquefaction stiffness for reliable estimation of reconsolidation strains is investigated. The importance of using an excess pore pressure ratio-dependent hydraulic conductivity to accurately model the pore pressure generation and dissipation patterns is analyzed. A new relationship between increase in hydraulic conductivity due to liquefaction and grain size diameter is proposed based on the results from this study and previous research. The importance of accurate estimation of relative density for reliable numerical predictions of post-liquefaction responses is also investigated. Overall, the displacements and settlements predicted by the numerical framework used in this study are within 50-200% of the corresponding experimental values, although this level of agreement can likely be improved through calibration of both dynamic and reconsolidation properties. This level of uncertainty is similar to those observed for other liquefaction problems by previous researchers, such as lateral spreading displacements observed in post-earthquake reconnaissance and centrifuge tests, and free-field settlements observed in numerical studies and comparisons with empirical relationships in Christchurch. Finally, the numerical methodology developed in this study is used to model a spatially variable soil deposit from Hollywood, South Carolina. The ability of the numerical framework in capturing the physical mechanisms involved in such a problem is investigated. The effects of various soil properties and input motion parameters on reconsolidation settlement are evaluated.